[LLD] [COFF] Allow embedded directives to be separated by null bytes

[mstorsjo]

Summary

The PE spec says that they will be separated by spaces, but link.exe handles it just fine if they are separated by null bytes as well.

Since these directives are parsed and split by TokenizeWindowsCommandLine, this function needs to learn to treat null bytes as separators - this patch makes that unconditional, but at worst we might need to make it an optional feature, only used when parsing .drectve.

[ruiu]

LGTM

lld/test/COFF/directives.s

+# REQUIRES: x86
+
+# RUN: llvm-mc -triple=x86_64-windows %s -filetype=obj -o %t.obj
+
+# RUN: lld-link -dll -out:%t.dll -entry:entry %t.obj -subsystem:console
+# RUN: llvm-objdump -p %t.dll | FileCheck %s
+
+# CHECK:      Export Table:
+# CHECK:      DLL name: directives.s.tmp.dll
+# CHECK:      Ordinal      RVA  Name
+# CHECK-NEXT:       0        0
+# CHECK-NEXT:       1   0x1000  exportfn1
+# CHECK-NEXT:       2   0x1000  exportfn2
+# CHECK-NEXT:       3   0x1000  exportfn3
+# CHECK-NEXT:       4   0x1000  exportfn4
+# CHECK-NEXT:       5   0x1000  exportfn5
+# CHECK-NEXT:       6   0x1000  exportfn6
+
+  .global entry
+  .global exportfn1
+  .global exportfn2
+  .global exportfn3
+  .global exportfn4
+  .global exportfn5
+  .global exportfn6
+  .text
+entry:
+exportfn1:
+exportfn2:
+exportfn3:
+exportfn4:
+exportfn5:
+exportfn6:
+  ret
+  .section .drectve
+# Test that directive strings can be separated by any combination of
+# spaces and null bytes.
+  .ascii "-export:exportfn1 "
+  .asciz "-export:exportfn2"
+  .asciz "-export:exportfn3"
+  .asciz "-export:exportfn4 "
+  .byte 0
+  .ascii " "
+  .byte 0
+  .asciz "-export:exportfn5"
+  .asciz " -export:exportfn6"

llvm/lib/Support/CommandLine.cpp

static bool EatsUnboundedNumberOfValues(const Option *O) {
   return O->getNumOccurrencesFlag() == cl::ZeroOrMore ||
          O->getNumOccurrencesFlag() == cl::OneOrMore;
 }
 
 static bool isWhitespace(char C) {
   return C == ' ' || C == '\t' || C == '\r' || C == '\n';
 }
 
+static bool isWhitespaceOrNull(char C) {
+  return isWhitespace(C) || C == '\0';
+}
+
 static bool isQuote(char C) { return C == '\"' || C == '\''; }
 
 void cl::TokenizeGNUCommandLine(StringRef Src, StringSaver &Saver,
                                 SmallVectorImpl<const char *> &NewArgv,
                                 bool MarkEOLs) {
   SmallString<128> Token;
   for (size_t I = 0, E = Src.size(); I != E; ++I) {
     // Consume runs of whitespace.
void cl::TokenizeWindowsCommandLine(StringRef Src, StringSaver &Saver,
   // end of the source string.
   enum { INIT, UNQUOTED, QUOTED } State = INIT;
   for (size_t I = 0, E = Src.size(); I != E; ++I) {
     char C = Src[I];
 
     // INIT state indicates that the current input index is at the start of
     // the string or between tokens.
     if (State == INIT) {
-      if (isWhitespace(C)) {
+      if (isWhitespaceOrNull(C)) {
         // Mark the end of lines in response files
         if (MarkEOLs && C == '\n')
           NewArgv.push_back(nullptr);
         continue;
       }
       if (C == '"') {
         State = QUOTED;
         continue;
       }
       if (C == '\\') {
         I = parseBackslash(Src, I, Token);
         State = UNQUOTED;
         continue;
       }
       Token.push_back(C);
       State = UNQUOTED;
       continue;
     }
 
     // UNQUOTED state means that it's reading a token not quoted by double
     // quotes.
     if (State == UNQUOTED) {
       // Whitespace means the end of the token.
-      if (isWhitespace(C)) {
+      if (isWhitespaceOrNull(C)) {
         NewArgv.push_back(Saver.save(StringRef(Token)).data());
         Token.clear();
         State = INIT;
         // Mark the end of lines in response files
         if (MarkEOLs && C == '\n')
           NewArgv.push_back(nullptr);
         continue;
       }